Eye detection systems based on an analysis of epipolar stereo images of the face have achieved a high degree of precision in reporting human eye location and gaze direction. A simple optical approach to implement this system is to use infrared (IR) light to illuminate the eyes. Light reflected off the inner posterior surface of the eye and is detected by an imager. Two sets of images are seen by an imager by using two sets of IR sources. The sources may be separated in time or have different wavelengths and the images may be separated by time-gating or wavelength filtering. These images can be processed digitally to yield bright circles indicative of pupil location.
A disadvantage of this approach is that is cannot be used when the subject is not close to the light source and imager. This is because more powerful IR light sources must be used. Such light sources pose a safety risk if the subject moves too close to the light source and gazes directly into them. Prolonged exposure to such high power signals could damage the eye. For example, if the system is designed to be safe at distances greater than 12 cm, then the approach does not work if the subject is more than 2 m away from the detector, as is the case for a television implementation for example.
Another approach uses a telephoto lens, passive illumination and conventional image processing, but this approach only works with appropriate ambient lighting and does not work through sunglasses. In addition, this requires substantial computational resources to locate an eye.